Um estudo sobre a aplicação de inteligência artificial em jogos

Abstract

Games are ideal environments for using Artificial Intelligence (AI) methods because there are well-defined mathematical models that help in implementing solutions for search problems. However, each problem must be analyzed individually, because there must be a specific modeling for data and information that will be used in the solution of an algorithm. To minimize the processing cost, scripts with actions are applied to games in order to make characters perform defined strategies in specific situations. This work studies the application of the Minimax algorithm for performing search in trees for solving decision problems in games. Seven prototypes were developed with different evaluation and search strategies in an implementation of the 2048 game. Performance results were analyzed in relation to runtime and score. TheMinimax algorithm solves the 2048 game, but in some cases the search had to be limited so the computer would not lock execution. This study shows that it is possible to apply AI to solve a game, but the tree search should be limited to avoid locks.